bwg-cli exposes 11 tools with basic schemas and descriptions, but falls short of production quality. Tool names lack consistent verb-noun structure (e.g., 'server ls' instead of 'list_servers'). Descriptions are present but generic (avg ~60 chars, well below the 194-char baseline). Parameter descriptions exist but are minimal. No output schemas documented. Error handling is absent from tool definitions. The server is a CLI wrapper around KiwiVM API; tool definitions appear inferred from Cobra command structure rather than explicitly registered with rich metadata. This limits per-tool scores to ~50 maximum per the rubric's inference penalty.

• Rename tools to verb_noun format: 'list_servers', 'create_server', 'update_server', 'delete_server', 'show_server', 'import_servers', 'verify_server', 'list_api_endpoints', 'call_api_endpoint', 'get_audit_log'. This makes intent clear to LLMs without reading descriptions.
• Document output schemas formally. For 'list_servers', define: {servers: [{name, veid, apiKey_masked, note, tags, sshUser, sshPort, fromEnv}], default: string, configPath: string}. For 'create_server', return {server: {name, veid, note, tags, sshUser, sshPort}, default: string}.
• Expand parameter descriptions to 50-200 chars with context. E.g., 'tag' → 'Tag for filtering servers in bulk operations (e.g., prod, staging, jp). Repeatable; replaces all existing tags when set.' Include valid formats and use cases.
• Add error handling guidance to tool descriptions. E.g., 'server add' → 'Adds a VPS to the fleet. Verifies credentials against KiwiVM API before saving. Returns error if VEID/key invalid or network unreachable; suggest --no-verify to skip check.'
• Explicitly register tools with MCP metadata (input schemas, output schemas, descriptions) rather than inferring from CLI. Use a tool registry or manifest to declare all 11 tools with complete JSON Schema definitions for inputs and outputs.